Merge tag 'loongarch-6.1' of git://git.kernel.org/pub/scm/linux/kernel/git/chenhuacai...
[platform/kernel/linux-starfive.git] / fs / ntfs3 / inode.c
1 // SPDX-License-Identifier: GPL-2.0
2 /*
3  *
4  * Copyright (C) 2019-2021 Paragon Software GmbH, All rights reserved.
5  *
6  */
7
8 #include <linux/buffer_head.h>
9 #include <linux/fs.h>
10 #include <linux/mpage.h>
11 #include <linux/namei.h>
12 #include <linux/nls.h>
13 #include <linux/uio.h>
14 #include <linux/writeback.h>
15
16 #include "debug.h"
17 #include "ntfs.h"
18 #include "ntfs_fs.h"
19
20 /*
21  * ntfs_read_mft - Read record and parses MFT.
22  */
23 static struct inode *ntfs_read_mft(struct inode *inode,
24                                    const struct cpu_str *name,
25                                    const struct MFT_REF *ref)
26 {
27         int err = 0;
28         struct ntfs_inode *ni = ntfs_i(inode);
29         struct super_block *sb = inode->i_sb;
30         struct ntfs_sb_info *sbi = sb->s_fs_info;
31         mode_t mode = 0;
32         struct ATTR_STD_INFO5 *std5 = NULL;
33         struct ATTR_LIST_ENTRY *le;
34         struct ATTRIB *attr;
35         bool is_match = false;
36         bool is_root = false;
37         bool is_dir;
38         unsigned long ino = inode->i_ino;
39         u32 rp_fa = 0, asize, t32;
40         u16 roff, rsize, names = 0;
41         const struct ATTR_FILE_NAME *fname = NULL;
42         const struct INDEX_ROOT *root;
43         struct REPARSE_DATA_BUFFER rp; // 0x18 bytes
44         u64 t64;
45         struct MFT_REC *rec;
46         struct runs_tree *run;
47
48         inode->i_op = NULL;
49         /* Setup 'uid' and 'gid' */
50         inode->i_uid = sbi->options->fs_uid;
51         inode->i_gid = sbi->options->fs_gid;
52
53         err = mi_init(&ni->mi, sbi, ino);
54         if (err)
55                 goto out;
56
57         if (!sbi->mft.ni && ino == MFT_REC_MFT && !sb->s_root) {
58                 t64 = sbi->mft.lbo >> sbi->cluster_bits;
59                 t32 = bytes_to_cluster(sbi, MFT_REC_VOL * sbi->record_size);
60                 sbi->mft.ni = ni;
61                 init_rwsem(&ni->file.run_lock);
62
63                 if (!run_add_entry(&ni->file.run, 0, t64, t32, true)) {
64                         err = -ENOMEM;
65                         goto out;
66                 }
67         }
68
69         err = mi_read(&ni->mi, ino == MFT_REC_MFT);
70
71         if (err)
72                 goto out;
73
74         rec = ni->mi.mrec;
75
76         if (sbi->flags & NTFS_FLAGS_LOG_REPLAYING) {
77                 ;
78         } else if (ref->seq != rec->seq) {
79                 err = -EINVAL;
80                 ntfs_err(sb, "MFT: r=%lx, expect seq=%x instead of %x!", ino,
81                          le16_to_cpu(ref->seq), le16_to_cpu(rec->seq));
82                 goto out;
83         } else if (!is_rec_inuse(rec)) {
84                 err = -EINVAL;
85                 ntfs_err(sb, "Inode r=%x is not in use!", (u32)ino);
86                 goto out;
87         }
88
89         if (le32_to_cpu(rec->total) != sbi->record_size) {
90                 /* Bad inode? */
91                 err = -EINVAL;
92                 goto out;
93         }
94
95         if (!is_rec_base(rec))
96                 goto Ok;
97
98         /* Record should contain $I30 root. */
99         is_dir = rec->flags & RECORD_FLAG_DIR;
100
101         inode->i_generation = le16_to_cpu(rec->seq);
102
103         /* Enumerate all struct Attributes MFT. */
104         le = NULL;
105         attr = NULL;
106
107         /*
108          * To reduce tab pressure use goto instead of
109          * while( (attr = ni_enum_attr_ex(ni, attr, &le, NULL) ))
110          */
111 next_attr:
112         run = NULL;
113         err = -EINVAL;
114         attr = ni_enum_attr_ex(ni, attr, &le, NULL);
115         if (!attr)
116                 goto end_enum;
117
118         if (le && le->vcn) {
119                 /* This is non primary attribute segment. Ignore if not MFT. */
120                 if (ino != MFT_REC_MFT || attr->type != ATTR_DATA)
121                         goto next_attr;
122
123                 run = &ni->file.run;
124                 asize = le32_to_cpu(attr->size);
125                 goto attr_unpack_run;
126         }
127
128         roff = attr->non_res ? 0 : le16_to_cpu(attr->res.data_off);
129         rsize = attr->non_res ? 0 : le32_to_cpu(attr->res.data_size);
130         asize = le32_to_cpu(attr->size);
131
132         switch (attr->type) {
133         case ATTR_STD:
134                 if (attr->non_res ||
135                     asize < sizeof(struct ATTR_STD_INFO) + roff ||
136                     rsize < sizeof(struct ATTR_STD_INFO))
137                         goto out;
138
139                 if (std5)
140                         goto next_attr;
141
142                 std5 = Add2Ptr(attr, roff);
143
144 #ifdef STATX_BTIME
145                 nt2kernel(std5->cr_time, &ni->i_crtime);
146 #endif
147                 nt2kernel(std5->a_time, &inode->i_atime);
148                 nt2kernel(std5->c_time, &inode->i_ctime);
149                 nt2kernel(std5->m_time, &inode->i_mtime);
150
151                 ni->std_fa = std5->fa;
152
153                 if (asize >= sizeof(struct ATTR_STD_INFO5) + roff &&
154                     rsize >= sizeof(struct ATTR_STD_INFO5))
155                         ni->std_security_id = std5->security_id;
156                 goto next_attr;
157
158         case ATTR_LIST:
159                 if (attr->name_len || le || ino == MFT_REC_LOG)
160                         goto out;
161
162                 err = ntfs_load_attr_list(ni, attr);
163                 if (err)
164                         goto out;
165
166                 le = NULL;
167                 attr = NULL;
168                 goto next_attr;
169
170         case ATTR_NAME:
171                 if (attr->non_res || asize < SIZEOF_ATTRIBUTE_FILENAME + roff ||
172                     rsize < SIZEOF_ATTRIBUTE_FILENAME)
173                         goto out;
174
175                 fname = Add2Ptr(attr, roff);
176                 if (fname->type == FILE_NAME_DOS)
177                         goto next_attr;
178
179                 names += 1;
180                 if (name && name->len == fname->name_len &&
181                     !ntfs_cmp_names_cpu(name, (struct le_str *)&fname->name_len,
182                                         NULL, false))
183                         is_match = true;
184
185                 goto next_attr;
186
187         case ATTR_DATA:
188                 if (is_dir) {
189                         /* Ignore data attribute in dir record. */
190                         goto next_attr;
191                 }
192
193                 if (ino == MFT_REC_BADCLUST && !attr->non_res)
194                         goto next_attr;
195
196                 if (attr->name_len &&
197                     ((ino != MFT_REC_BADCLUST || !attr->non_res ||
198                       attr->name_len != ARRAY_SIZE(BAD_NAME) ||
199                       memcmp(attr_name(attr), BAD_NAME, sizeof(BAD_NAME))) &&
200                      (ino != MFT_REC_SECURE || !attr->non_res ||
201                       attr->name_len != ARRAY_SIZE(SDS_NAME) ||
202                       memcmp(attr_name(attr), SDS_NAME, sizeof(SDS_NAME))))) {
203                         /* File contains stream attribute. Ignore it. */
204                         goto next_attr;
205                 }
206
207                 if (is_attr_sparsed(attr))
208                         ni->std_fa |= FILE_ATTRIBUTE_SPARSE_FILE;
209                 else
210                         ni->std_fa &= ~FILE_ATTRIBUTE_SPARSE_FILE;
211
212                 if (is_attr_compressed(attr))
213                         ni->std_fa |= FILE_ATTRIBUTE_COMPRESSED;
214                 else
215                         ni->std_fa &= ~FILE_ATTRIBUTE_COMPRESSED;
216
217                 if (is_attr_encrypted(attr))
218                         ni->std_fa |= FILE_ATTRIBUTE_ENCRYPTED;
219                 else
220                         ni->std_fa &= ~FILE_ATTRIBUTE_ENCRYPTED;
221
222                 if (!attr->non_res) {
223                         ni->i_valid = inode->i_size = rsize;
224                         inode_set_bytes(inode, rsize);
225                 }
226
227                 mode = S_IFREG | (0777 & sbi->options->fs_fmask_inv);
228
229                 if (!attr->non_res) {
230                         ni->ni_flags |= NI_FLAG_RESIDENT;
231                         goto next_attr;
232                 }
233
234                 inode_set_bytes(inode, attr_ondisk_size(attr));
235
236                 ni->i_valid = le64_to_cpu(attr->nres.valid_size);
237                 inode->i_size = le64_to_cpu(attr->nres.data_size);
238                 if (!attr->nres.alloc_size)
239                         goto next_attr;
240
241                 run = ino == MFT_REC_BITMAP ? &sbi->used.bitmap.run
242                                             : &ni->file.run;
243                 break;
244
245         case ATTR_ROOT:
246                 if (attr->non_res)
247                         goto out;
248
249                 root = Add2Ptr(attr, roff);
250                 is_root = true;
251
252                 if (attr->name_len != ARRAY_SIZE(I30_NAME) ||
253                     memcmp(attr_name(attr), I30_NAME, sizeof(I30_NAME)))
254                         goto next_attr;
255
256                 if (root->type != ATTR_NAME ||
257                     root->rule != NTFS_COLLATION_TYPE_FILENAME)
258                         goto out;
259
260                 if (!is_dir)
261                         goto next_attr;
262
263                 ni->ni_flags |= NI_FLAG_DIR;
264
265                 err = indx_init(&ni->dir, sbi, attr, INDEX_MUTEX_I30);
266                 if (err)
267                         goto out;
268
269                 mode = sb->s_root
270                                ? (S_IFDIR | (0777 & sbi->options->fs_dmask_inv))
271                                : (S_IFDIR | 0777);
272                 goto next_attr;
273
274         case ATTR_ALLOC:
275                 if (!is_root || attr->name_len != ARRAY_SIZE(I30_NAME) ||
276                     memcmp(attr_name(attr), I30_NAME, sizeof(I30_NAME)))
277                         goto next_attr;
278
279                 inode->i_size = le64_to_cpu(attr->nres.data_size);
280                 ni->i_valid = le64_to_cpu(attr->nres.valid_size);
281                 inode_set_bytes(inode, le64_to_cpu(attr->nres.alloc_size));
282
283                 run = &ni->dir.alloc_run;
284                 break;
285
286         case ATTR_BITMAP:
287                 if (ino == MFT_REC_MFT) {
288                         if (!attr->non_res)
289                                 goto out;
290 #ifndef CONFIG_NTFS3_64BIT_CLUSTER
291                         /* 0x20000000 = 2^32 / 8 */
292                         if (le64_to_cpu(attr->nres.alloc_size) >= 0x20000000)
293                                 goto out;
294 #endif
295                         run = &sbi->mft.bitmap.run;
296                         break;
297                 } else if (is_dir && attr->name_len == ARRAY_SIZE(I30_NAME) &&
298                            !memcmp(attr_name(attr), I30_NAME,
299                                    sizeof(I30_NAME)) &&
300                            attr->non_res) {
301                         run = &ni->dir.bitmap_run;
302                         break;
303                 }
304                 goto next_attr;
305
306         case ATTR_REPARSE:
307                 if (attr->name_len)
308                         goto next_attr;
309
310                 rp_fa = ni_parse_reparse(ni, attr, &rp);
311                 switch (rp_fa) {
312                 case REPARSE_LINK:
313                         /*
314                          * Normal symlink.
315                          * Assume one unicode symbol == one utf8.
316                          */
317                         inode->i_size = le16_to_cpu(rp.SymbolicLinkReparseBuffer
318                                                             .PrintNameLength) /
319                                         sizeof(u16);
320
321                         ni->i_valid = inode->i_size;
322
323                         /* Clear directory bit. */
324                         if (ni->ni_flags & NI_FLAG_DIR) {
325                                 indx_clear(&ni->dir);
326                                 memset(&ni->dir, 0, sizeof(ni->dir));
327                                 ni->ni_flags &= ~NI_FLAG_DIR;
328                         } else {
329                                 run_close(&ni->file.run);
330                         }
331                         mode = S_IFLNK | 0777;
332                         is_dir = false;
333                         if (attr->non_res) {
334                                 run = &ni->file.run;
335                                 goto attr_unpack_run; // Double break.
336                         }
337                         break;
338
339                 case REPARSE_COMPRESSED:
340                         break;
341
342                 case REPARSE_DEDUPLICATED:
343                         break;
344                 }
345                 goto next_attr;
346
347         case ATTR_EA_INFO:
348                 if (!attr->name_len &&
349                     resident_data_ex(attr, sizeof(struct EA_INFO))) {
350                         ni->ni_flags |= NI_FLAG_EA;
351                         /*
352                          * ntfs_get_wsl_perm updates inode->i_uid, inode->i_gid, inode->i_mode
353                          */
354                         inode->i_mode = mode;
355                         ntfs_get_wsl_perm(inode);
356                         mode = inode->i_mode;
357                 }
358                 goto next_attr;
359
360         default:
361                 goto next_attr;
362         }
363
364 attr_unpack_run:
365         roff = le16_to_cpu(attr->nres.run_off);
366
367         t64 = le64_to_cpu(attr->nres.svcn);
368         err = run_unpack_ex(run, sbi, ino, t64, le64_to_cpu(attr->nres.evcn),
369                             t64, Add2Ptr(attr, roff), asize - roff);
370         if (err < 0)
371                 goto out;
372         err = 0;
373         goto next_attr;
374
375 end_enum:
376
377         if (!std5)
378                 goto out;
379
380         if (!is_match && name) {
381                 /* Reuse rec as buffer for ascii name. */
382                 err = -ENOENT;
383                 goto out;
384         }
385
386         if (std5->fa & FILE_ATTRIBUTE_READONLY)
387                 mode &= ~0222;
388
389         if (!names) {
390                 err = -EINVAL;
391                 goto out;
392         }
393
394         if (names != le16_to_cpu(rec->hard_links)) {
395                 /* Correct minor error on the fly. Do not mark inode as dirty. */
396                 rec->hard_links = cpu_to_le16(names);
397                 ni->mi.dirty = true;
398         }
399
400         set_nlink(inode, names);
401
402         if (S_ISDIR(mode)) {
403                 ni->std_fa |= FILE_ATTRIBUTE_DIRECTORY;
404
405                 /*
406                  * Dot and dot-dot should be included in count but was not
407                  * included in enumeration.
408                  * Usually a hard links to directories are disabled.
409                  */
410                 inode->i_op = &ntfs_dir_inode_operations;
411                 inode->i_fop = &ntfs_dir_operations;
412                 ni->i_valid = 0;
413         } else if (S_ISLNK(mode)) {
414                 ni->std_fa &= ~FILE_ATTRIBUTE_DIRECTORY;
415                 inode->i_op = &ntfs_link_inode_operations;
416                 inode->i_fop = NULL;
417                 inode_nohighmem(inode);
418         } else if (S_ISREG(mode)) {
419                 ni->std_fa &= ~FILE_ATTRIBUTE_DIRECTORY;
420                 inode->i_op = &ntfs_file_inode_operations;
421                 inode->i_fop = &ntfs_file_operations;
422                 inode->i_mapping->a_ops =
423                         is_compressed(ni) ? &ntfs_aops_cmpr : &ntfs_aops;
424                 if (ino != MFT_REC_MFT)
425                         init_rwsem(&ni->file.run_lock);
426         } else if (S_ISCHR(mode) || S_ISBLK(mode) || S_ISFIFO(mode) ||
427                    S_ISSOCK(mode)) {
428                 inode->i_op = &ntfs_special_inode_operations;
429                 init_special_inode(inode, mode, inode->i_rdev);
430         } else if (fname && fname->home.low == cpu_to_le32(MFT_REC_EXTEND) &&
431                    fname->home.seq == cpu_to_le16(MFT_REC_EXTEND)) {
432                 /* Records in $Extend are not a files or general directories. */
433                 inode->i_op = &ntfs_file_inode_operations;
434         } else {
435                 err = -EINVAL;
436                 goto out;
437         }
438
439         if ((sbi->options->sys_immutable &&
440              (std5->fa & FILE_ATTRIBUTE_SYSTEM)) &&
441             !S_ISFIFO(mode) && !S_ISSOCK(mode) && !S_ISLNK(mode)) {
442                 inode->i_flags |= S_IMMUTABLE;
443         } else {
444                 inode->i_flags &= ~S_IMMUTABLE;
445         }
446
447         inode->i_mode = mode;
448         if (!(ni->ni_flags & NI_FLAG_EA)) {
449                 /* If no xattr then no security (stored in xattr). */
450                 inode->i_flags |= S_NOSEC;
451         }
452
453 Ok:
454         if (ino == MFT_REC_MFT && !sb->s_root)
455                 sbi->mft.ni = NULL;
456
457         unlock_new_inode(inode);
458
459         return inode;
460
461 out:
462         if (ino == MFT_REC_MFT && !sb->s_root)
463                 sbi->mft.ni = NULL;
464
465         iget_failed(inode);
466         return ERR_PTR(err);
467 }
468
469 /*
470  * ntfs_test_inode
471  *
472  * Return: 1 if match.
473  */
474 static int ntfs_test_inode(struct inode *inode, void *data)
475 {
476         struct MFT_REF *ref = data;
477
478         return ino_get(ref) == inode->i_ino;
479 }
480
481 static int ntfs_set_inode(struct inode *inode, void *data)
482 {
483         const struct MFT_REF *ref = data;
484
485         inode->i_ino = ino_get(ref);
486         return 0;
487 }
488
489 struct inode *ntfs_iget5(struct super_block *sb, const struct MFT_REF *ref,
490                          const struct cpu_str *name)
491 {
492         struct inode *inode;
493
494         inode = iget5_locked(sb, ino_get(ref), ntfs_test_inode, ntfs_set_inode,
495                              (void *)ref);
496         if (unlikely(!inode))
497                 return ERR_PTR(-ENOMEM);
498
499         /* If this is a freshly allocated inode, need to read it now. */
500         if (inode->i_state & I_NEW)
501                 inode = ntfs_read_mft(inode, name, ref);
502         else if (ref->seq != ntfs_i(inode)->mi.mrec->seq) {
503                 /* Inode overlaps? */
504                 _ntfs_bad_inode(inode);
505         }
506
507         return inode;
508 }
509
510 enum get_block_ctx {
511         GET_BLOCK_GENERAL = 0,
512         GET_BLOCK_WRITE_BEGIN = 1,
513         GET_BLOCK_DIRECT_IO_R = 2,
514         GET_BLOCK_DIRECT_IO_W = 3,
515         GET_BLOCK_BMAP = 4,
516 };
517
518 static noinline int ntfs_get_block_vbo(struct inode *inode, u64 vbo,
519                                        struct buffer_head *bh, int create,
520                                        enum get_block_ctx ctx)
521 {
522         struct super_block *sb = inode->i_sb;
523         struct ntfs_sb_info *sbi = sb->s_fs_info;
524         struct ntfs_inode *ni = ntfs_i(inode);
525         struct page *page = bh->b_page;
526         u8 cluster_bits = sbi->cluster_bits;
527         u32 block_size = sb->s_blocksize;
528         u64 bytes, lbo, valid;
529         u32 off;
530         int err;
531         CLST vcn, lcn, len;
532         bool new;
533
534         /* Clear previous state. */
535         clear_buffer_new(bh);
536         clear_buffer_uptodate(bh);
537
538         /* Direct write uses 'create=0'. */
539         if (!create && vbo >= ni->i_valid) {
540                 /* Out of valid. */
541                 return 0;
542         }
543
544         if (vbo >= inode->i_size) {
545                 /* Out of size. */
546                 return 0;
547         }
548
549         if (is_resident(ni)) {
550                 ni_lock(ni);
551                 err = attr_data_read_resident(ni, page);
552                 ni_unlock(ni);
553
554                 if (!err)
555                         set_buffer_uptodate(bh);
556                 bh->b_size = block_size;
557                 return err;
558         }
559
560         vcn = vbo >> cluster_bits;
561         off = vbo & sbi->cluster_mask;
562         new = false;
563
564         err = attr_data_get_block(ni, vcn, 1, &lcn, &len, create ? &new : NULL);
565         if (err)
566                 goto out;
567
568         if (!len)
569                 return 0;
570
571         bytes = ((u64)len << cluster_bits) - off;
572
573         if (lcn == SPARSE_LCN) {
574                 if (!create) {
575                         if (bh->b_size > bytes)
576                                 bh->b_size = bytes;
577                         return 0;
578                 }
579                 WARN_ON(1);
580         }
581
582         if (new) {
583                 set_buffer_new(bh);
584                 if ((len << cluster_bits) > block_size)
585                         ntfs_sparse_cluster(inode, page, vcn, len);
586         }
587
588         lbo = ((u64)lcn << cluster_bits) + off;
589
590         set_buffer_mapped(bh);
591         bh->b_bdev = sb->s_bdev;
592         bh->b_blocknr = lbo >> sb->s_blocksize_bits;
593
594         valid = ni->i_valid;
595
596         if (ctx == GET_BLOCK_DIRECT_IO_W) {
597                 /* ntfs_direct_IO will update ni->i_valid. */
598                 if (vbo >= valid)
599                         set_buffer_new(bh);
600         } else if (create) {
601                 /* Normal write. */
602                 if (bytes > bh->b_size)
603                         bytes = bh->b_size;
604
605                 if (vbo >= valid)
606                         set_buffer_new(bh);
607
608                 if (vbo + bytes > valid) {
609                         ni->i_valid = vbo + bytes;
610                         mark_inode_dirty(inode);
611                 }
612         } else if (vbo >= valid) {
613                 /* Read out of valid data. */
614                 /* Should never be here 'cause already checked. */
615                 clear_buffer_mapped(bh);
616         } else if (vbo + bytes <= valid) {
617                 /* Normal read. */
618         } else if (vbo + block_size <= valid) {
619                 /* Normal short read. */
620                 bytes = block_size;
621         } else {
622                 /*
623                  * Read across valid size: vbo < valid && valid < vbo + block_size
624                  */
625                 bytes = block_size;
626
627                 if (page) {
628                         u32 voff = valid - vbo;
629
630                         bh->b_size = block_size;
631                         off = vbo & (PAGE_SIZE - 1);
632                         set_bh_page(bh, page, off);
633                         err = bh_read(bh, 0);
634                         if (err < 0)
635                                 goto out;
636                         zero_user_segment(page, off + voff, off + block_size);
637                 }
638         }
639
640         if (bh->b_size > bytes)
641                 bh->b_size = bytes;
642
643 #ifndef __LP64__
644         if (ctx == GET_BLOCK_DIRECT_IO_W || ctx == GET_BLOCK_DIRECT_IO_R) {
645                 static_assert(sizeof(size_t) < sizeof(loff_t));
646                 if (bytes > 0x40000000u)
647                         bh->b_size = 0x40000000u;
648         }
649 #endif
650
651         return 0;
652
653 out:
654         return err;
655 }
656
657 int ntfs_get_block(struct inode *inode, sector_t vbn,
658                    struct buffer_head *bh_result, int create)
659 {
660         return ntfs_get_block_vbo(inode, (u64)vbn << inode->i_blkbits,
661                                   bh_result, create, GET_BLOCK_GENERAL);
662 }
663
664 static int ntfs_get_block_bmap(struct inode *inode, sector_t vsn,
665                                struct buffer_head *bh_result, int create)
666 {
667         return ntfs_get_block_vbo(inode,
668                                   (u64)vsn << inode->i_sb->s_blocksize_bits,
669                                   bh_result, create, GET_BLOCK_BMAP);
670 }
671
672 static sector_t ntfs_bmap(struct address_space *mapping, sector_t block)
673 {
674         return generic_block_bmap(mapping, block, ntfs_get_block_bmap);
675 }
676
677 static int ntfs_read_folio(struct file *file, struct folio *folio)
678 {
679         struct page *page = &folio->page;
680         int err;
681         struct address_space *mapping = page->mapping;
682         struct inode *inode = mapping->host;
683         struct ntfs_inode *ni = ntfs_i(inode);
684
685         if (is_resident(ni)) {
686                 ni_lock(ni);
687                 err = attr_data_read_resident(ni, page);
688                 ni_unlock(ni);
689                 if (err != E_NTFS_NONRESIDENT) {
690                         unlock_page(page);
691                         return err;
692                 }
693         }
694
695         if (is_compressed(ni)) {
696                 ni_lock(ni);
697                 err = ni_readpage_cmpr(ni, page);
698                 ni_unlock(ni);
699                 return err;
700         }
701
702         /* Normal + sparse files. */
703         return mpage_read_folio(folio, ntfs_get_block);
704 }
705
706 static void ntfs_readahead(struct readahead_control *rac)
707 {
708         struct address_space *mapping = rac->mapping;
709         struct inode *inode = mapping->host;
710         struct ntfs_inode *ni = ntfs_i(inode);
711         u64 valid;
712         loff_t pos;
713
714         if (is_resident(ni)) {
715                 /* No readahead for resident. */
716                 return;
717         }
718
719         if (is_compressed(ni)) {
720                 /* No readahead for compressed. */
721                 return;
722         }
723
724         valid = ni->i_valid;
725         pos = readahead_pos(rac);
726
727         if (valid < i_size_read(inode) && pos <= valid &&
728             valid < pos + readahead_length(rac)) {
729                 /* Range cross 'valid'. Read it page by page. */
730                 return;
731         }
732
733         mpage_readahead(rac, ntfs_get_block);
734 }
735
736 static int ntfs_get_block_direct_IO_R(struct inode *inode, sector_t iblock,
737                                       struct buffer_head *bh_result, int create)
738 {
739         return ntfs_get_block_vbo(inode, (u64)iblock << inode->i_blkbits,
740                                   bh_result, create, GET_BLOCK_DIRECT_IO_R);
741 }
742
743 static int ntfs_get_block_direct_IO_W(struct inode *inode, sector_t iblock,
744                                       struct buffer_head *bh_result, int create)
745 {
746         return ntfs_get_block_vbo(inode, (u64)iblock << inode->i_blkbits,
747                                   bh_result, create, GET_BLOCK_DIRECT_IO_W);
748 }
749
750 static ssize_t ntfs_direct_IO(struct kiocb *iocb, struct iov_iter *iter)
751 {
752         struct file *file = iocb->ki_filp;
753         struct address_space *mapping = file->f_mapping;
754         struct inode *inode = mapping->host;
755         struct ntfs_inode *ni = ntfs_i(inode);
756         loff_t vbo = iocb->ki_pos;
757         loff_t end;
758         int wr = iov_iter_rw(iter) & WRITE;
759         size_t iter_count = iov_iter_count(iter);
760         loff_t valid;
761         ssize_t ret;
762
763         if (is_resident(ni)) {
764                 /* Switch to buffered write. */
765                 ret = 0;
766                 goto out;
767         }
768
769         ret = blockdev_direct_IO(iocb, inode, iter,
770                                  wr ? ntfs_get_block_direct_IO_W
771                                     : ntfs_get_block_direct_IO_R);
772
773         if (ret > 0)
774                 end = vbo + ret;
775         else if (wr && ret == -EIOCBQUEUED)
776                 end = vbo + iter_count;
777         else
778                 goto out;
779
780         valid = ni->i_valid;
781         if (wr) {
782                 if (end > valid && !S_ISBLK(inode->i_mode)) {
783                         ni->i_valid = end;
784                         mark_inode_dirty(inode);
785                 }
786         } else if (vbo < valid && valid < end) {
787                 /* Fix page. */
788                 iov_iter_revert(iter, end - valid);
789                 iov_iter_zero(end - valid, iter);
790         }
791
792 out:
793         return ret;
794 }
795
796 int ntfs_set_size(struct inode *inode, u64 new_size)
797 {
798         struct super_block *sb = inode->i_sb;
799         struct ntfs_sb_info *sbi = sb->s_fs_info;
800         struct ntfs_inode *ni = ntfs_i(inode);
801         int err;
802
803         /* Check for maximum file size. */
804         if (is_sparsed(ni) || is_compressed(ni)) {
805                 if (new_size > sbi->maxbytes_sparse) {
806                         err = -EFBIG;
807                         goto out;
808                 }
809         } else if (new_size > sbi->maxbytes) {
810                 err = -EFBIG;
811                 goto out;
812         }
813
814         ni_lock(ni);
815         down_write(&ni->file.run_lock);
816
817         err = attr_set_size(ni, ATTR_DATA, NULL, 0, &ni->file.run, new_size,
818                             &ni->i_valid, true, NULL);
819
820         up_write(&ni->file.run_lock);
821         ni_unlock(ni);
822
823         mark_inode_dirty(inode);
824
825 out:
826         return err;
827 }
828
829 static int ntfs_writepage(struct page *page, struct writeback_control *wbc)
830 {
831         struct address_space *mapping = page->mapping;
832         struct inode *inode = mapping->host;
833         struct ntfs_inode *ni = ntfs_i(inode);
834         int err;
835
836         if (is_resident(ni)) {
837                 ni_lock(ni);
838                 err = attr_data_write_resident(ni, page);
839                 ni_unlock(ni);
840                 if (err != E_NTFS_NONRESIDENT) {
841                         unlock_page(page);
842                         return err;
843                 }
844         }
845
846         return block_write_full_page(page, ntfs_get_block, wbc);
847 }
848
849 static int ntfs_writepages(struct address_space *mapping,
850                            struct writeback_control *wbc)
851 {
852         /* Redirect call to 'ntfs_writepage' for resident files. */
853         if (is_resident(ntfs_i(mapping->host)))
854                 return generic_writepages(mapping, wbc);
855         return mpage_writepages(mapping, wbc, ntfs_get_block);
856 }
857
858 static int ntfs_get_block_write_begin(struct inode *inode, sector_t vbn,
859                                       struct buffer_head *bh_result, int create)
860 {
861         return ntfs_get_block_vbo(inode, (u64)vbn << inode->i_blkbits,
862                                   bh_result, create, GET_BLOCK_WRITE_BEGIN);
863 }
864
865 int ntfs_write_begin(struct file *file, struct address_space *mapping,
866                      loff_t pos, u32 len, struct page **pagep, void **fsdata)
867 {
868         int err;
869         struct inode *inode = mapping->host;
870         struct ntfs_inode *ni = ntfs_i(inode);
871
872         *pagep = NULL;
873         if (is_resident(ni)) {
874                 struct page *page = grab_cache_page_write_begin(
875                         mapping, pos >> PAGE_SHIFT);
876
877                 if (!page) {
878                         err = -ENOMEM;
879                         goto out;
880                 }
881
882                 ni_lock(ni);
883                 err = attr_data_read_resident(ni, page);
884                 ni_unlock(ni);
885
886                 if (!err) {
887                         *pagep = page;
888                         goto out;
889                 }
890                 unlock_page(page);
891                 put_page(page);
892
893                 if (err != E_NTFS_NONRESIDENT)
894                         goto out;
895         }
896
897         err = block_write_begin(mapping, pos, len, pagep,
898                                 ntfs_get_block_write_begin);
899
900 out:
901         return err;
902 }
903
904 /*
905  * ntfs_write_end - Address_space_operations::write_end.
906  */
907 int ntfs_write_end(struct file *file, struct address_space *mapping,
908                    loff_t pos, u32 len, u32 copied, struct page *page,
909                    void *fsdata)
910 {
911         struct inode *inode = mapping->host;
912         struct ntfs_inode *ni = ntfs_i(inode);
913         u64 valid = ni->i_valid;
914         bool dirty = false;
915         int err;
916
917         if (is_resident(ni)) {
918                 ni_lock(ni);
919                 err = attr_data_write_resident(ni, page);
920                 ni_unlock(ni);
921                 if (!err) {
922                         dirty = true;
923                         /* Clear any buffers in page. */
924                         if (page_has_buffers(page)) {
925                                 struct buffer_head *head, *bh;
926
927                                 bh = head = page_buffers(page);
928                                 do {
929                                         clear_buffer_dirty(bh);
930                                         clear_buffer_mapped(bh);
931                                         set_buffer_uptodate(bh);
932                                 } while (head != (bh = bh->b_this_page));
933                         }
934                         SetPageUptodate(page);
935                         err = copied;
936                 }
937                 unlock_page(page);
938                 put_page(page);
939         } else {
940                 err = generic_write_end(file, mapping, pos, len, copied, page,
941                                         fsdata);
942         }
943
944         if (err >= 0) {
945                 if (!(ni->std_fa & FILE_ATTRIBUTE_ARCHIVE)) {
946                         inode->i_ctime = inode->i_mtime = current_time(inode);
947                         ni->std_fa |= FILE_ATTRIBUTE_ARCHIVE;
948                         dirty = true;
949                 }
950
951                 if (valid != ni->i_valid) {
952                         /* ni->i_valid is changed in ntfs_get_block_vbo. */
953                         dirty = true;
954                 }
955
956                 if (dirty)
957                         mark_inode_dirty(inode);
958         }
959
960         return err;
961 }
962
963 int reset_log_file(struct inode *inode)
964 {
965         int err;
966         loff_t pos = 0;
967         u32 log_size = inode->i_size;
968         struct address_space *mapping = inode->i_mapping;
969
970         for (;;) {
971                 u32 len;
972                 void *kaddr;
973                 struct page *page;
974
975                 len = pos + PAGE_SIZE > log_size ? (log_size - pos) : PAGE_SIZE;
976
977                 err = block_write_begin(mapping, pos, len, &page,
978                                         ntfs_get_block_write_begin);
979                 if (err)
980                         goto out;
981
982                 kaddr = kmap_atomic(page);
983                 memset(kaddr, -1, len);
984                 kunmap_atomic(kaddr);
985                 flush_dcache_page(page);
986
987                 err = block_write_end(NULL, mapping, pos, len, len, page, NULL);
988                 if (err < 0)
989                         goto out;
990                 pos += len;
991
992                 if (pos >= log_size)
993                         break;
994                 balance_dirty_pages_ratelimited(mapping);
995         }
996 out:
997         mark_inode_dirty_sync(inode);
998
999         return err;
1000 }
1001
1002 int ntfs3_write_inode(struct inode *inode, struct writeback_control *wbc)
1003 {
1004         return _ni_write_inode(inode, wbc->sync_mode == WB_SYNC_ALL);
1005 }
1006
1007 int ntfs_sync_inode(struct inode *inode)
1008 {
1009         return _ni_write_inode(inode, 1);
1010 }
1011
1012 /*
1013  * writeback_inode - Helper function for ntfs_flush_inodes().
1014  *
1015  * This writes both the inode and the file data blocks, waiting
1016  * for in flight data blocks before the start of the call.  It
1017  * does not wait for any io started during the call.
1018  */
1019 static int writeback_inode(struct inode *inode)
1020 {
1021         int ret = sync_inode_metadata(inode, 0);
1022
1023         if (!ret)
1024                 ret = filemap_fdatawrite(inode->i_mapping);
1025         return ret;
1026 }
1027
1028 /*
1029  * ntfs_flush_inodes
1030  *
1031  * Write data and metadata corresponding to i1 and i2.  The io is
1032  * started but we do not wait for any of it to finish.
1033  *
1034  * filemap_flush() is used for the block device, so if there is a dirty
1035  * page for a block already in flight, we will not wait and start the
1036  * io over again.
1037  */
1038 int ntfs_flush_inodes(struct super_block *sb, struct inode *i1,
1039                       struct inode *i2)
1040 {
1041         int ret = 0;
1042
1043         if (i1)
1044                 ret = writeback_inode(i1);
1045         if (!ret && i2)
1046                 ret = writeback_inode(i2);
1047         if (!ret)
1048                 ret = sync_blockdev_nowait(sb->s_bdev);
1049         return ret;
1050 }
1051
1052 int inode_write_data(struct inode *inode, const void *data, size_t bytes)
1053 {
1054         pgoff_t idx;
1055
1056         /* Write non resident data. */
1057         for (idx = 0; bytes; idx++) {
1058                 size_t op = bytes > PAGE_SIZE ? PAGE_SIZE : bytes;
1059                 struct page *page = ntfs_map_page(inode->i_mapping, idx);
1060
1061                 if (IS_ERR(page))
1062                         return PTR_ERR(page);
1063
1064                 lock_page(page);
1065                 WARN_ON(!PageUptodate(page));
1066                 ClearPageUptodate(page);
1067
1068                 memcpy(page_address(page), data, op);
1069
1070                 flush_dcache_page(page);
1071                 SetPageUptodate(page);
1072                 unlock_page(page);
1073
1074                 ntfs_unmap_page(page);
1075
1076                 bytes -= op;
1077                 data = Add2Ptr(data, PAGE_SIZE);
1078         }
1079         return 0;
1080 }
1081
1082 /*
1083  * ntfs_reparse_bytes
1084  *
1085  * Number of bytes for REPARSE_DATA_BUFFER(IO_REPARSE_TAG_SYMLINK)
1086  * for unicode string of @uni_len length.
1087  */
1088 static inline u32 ntfs_reparse_bytes(u32 uni_len)
1089 {
1090         /* Header + unicode string + decorated unicode string. */
1091         return sizeof(short) * (2 * uni_len + 4) +
1092                offsetof(struct REPARSE_DATA_BUFFER,
1093                         SymbolicLinkReparseBuffer.PathBuffer);
1094 }
1095
1096 static struct REPARSE_DATA_BUFFER *
1097 ntfs_create_reparse_buffer(struct ntfs_sb_info *sbi, const char *symname,
1098                            u32 size, u16 *nsize)
1099 {
1100         int i, err;
1101         struct REPARSE_DATA_BUFFER *rp;
1102         __le16 *rp_name;
1103         typeof(rp->SymbolicLinkReparseBuffer) *rs;
1104
1105         rp = kzalloc(ntfs_reparse_bytes(2 * size + 2), GFP_NOFS);
1106         if (!rp)
1107                 return ERR_PTR(-ENOMEM);
1108
1109         rs = &rp->SymbolicLinkReparseBuffer;
1110         rp_name = rs->PathBuffer;
1111
1112         /* Convert link name to UTF-16. */
1113         err = ntfs_nls_to_utf16(sbi, symname, size,
1114                                 (struct cpu_str *)(rp_name - 1), 2 * size,
1115                                 UTF16_LITTLE_ENDIAN);
1116         if (err < 0)
1117                 goto out;
1118
1119         /* err = the length of unicode name of symlink. */
1120         *nsize = ntfs_reparse_bytes(err);
1121
1122         if (*nsize > sbi->reparse.max_size) {
1123                 err = -EFBIG;
1124                 goto out;
1125         }
1126
1127         /* Translate Linux '/' into Windows '\'. */
1128         for (i = 0; i < err; i++) {
1129                 if (rp_name[i] == cpu_to_le16('/'))
1130                         rp_name[i] = cpu_to_le16('\\');
1131         }
1132
1133         rp->ReparseTag = IO_REPARSE_TAG_SYMLINK;
1134         rp->ReparseDataLength =
1135                 cpu_to_le16(*nsize - offsetof(struct REPARSE_DATA_BUFFER,
1136                                               SymbolicLinkReparseBuffer));
1137
1138         /* PrintName + SubstituteName. */
1139         rs->SubstituteNameOffset = cpu_to_le16(sizeof(short) * err);
1140         rs->SubstituteNameLength = cpu_to_le16(sizeof(short) * err + 8);
1141         rs->PrintNameLength = rs->SubstituteNameOffset;
1142
1143         /*
1144          * TODO: Use relative path if possible to allow Windows to
1145          * parse this path.
1146          * 0-absolute path 1- relative path (SYMLINK_FLAG_RELATIVE).
1147          */
1148         rs->Flags = 0;
1149
1150         memmove(rp_name + err + 4, rp_name, sizeof(short) * err);
1151
1152         /* Decorate SubstituteName. */
1153         rp_name += err;
1154         rp_name[0] = cpu_to_le16('\\');
1155         rp_name[1] = cpu_to_le16('?');
1156         rp_name[2] = cpu_to_le16('?');
1157         rp_name[3] = cpu_to_le16('\\');
1158
1159         return rp;
1160 out:
1161         kfree(rp);
1162         return ERR_PTR(err);
1163 }
1164
1165 struct inode *ntfs_create_inode(struct user_namespace *mnt_userns,
1166                                 struct inode *dir, struct dentry *dentry,
1167                                 const struct cpu_str *uni, umode_t mode,
1168                                 dev_t dev, const char *symname, u32 size,
1169                                 struct ntfs_fnd *fnd)
1170 {
1171         int err;
1172         struct super_block *sb = dir->i_sb;
1173         struct ntfs_sb_info *sbi = sb->s_fs_info;
1174         const struct qstr *name = &dentry->d_name;
1175         CLST ino = 0;
1176         struct ntfs_inode *dir_ni = ntfs_i(dir);
1177         struct ntfs_inode *ni = NULL;
1178         struct inode *inode = NULL;
1179         struct ATTRIB *attr;
1180         struct ATTR_STD_INFO5 *std5;
1181         struct ATTR_FILE_NAME *fname;
1182         struct MFT_REC *rec;
1183         u32 asize, dsize, sd_size;
1184         enum FILE_ATTRIBUTE fa;
1185         __le32 security_id = SECURITY_ID_INVALID;
1186         CLST vcn;
1187         const void *sd;
1188         u16 t16, nsize = 0, aid = 0;
1189         struct INDEX_ROOT *root, *dir_root;
1190         struct NTFS_DE *e, *new_de = NULL;
1191         struct REPARSE_DATA_BUFFER *rp = NULL;
1192         bool rp_inserted = false;
1193
1194         ni_lock_dir(dir_ni);
1195
1196         dir_root = indx_get_root(&dir_ni->dir, dir_ni, NULL, NULL);
1197         if (!dir_root) {
1198                 err = -EINVAL;
1199                 goto out1;
1200         }
1201
1202         if (S_ISDIR(mode)) {
1203                 /* Use parent's directory attributes. */
1204                 fa = dir_ni->std_fa | FILE_ATTRIBUTE_DIRECTORY |
1205                      FILE_ATTRIBUTE_ARCHIVE;
1206                 /*
1207                  * By default child directory inherits parent attributes.
1208                  * Root directory is hidden + system.
1209                  * Make an exception for children in root.
1210                  */
1211                 if (dir->i_ino == MFT_REC_ROOT)
1212                         fa &= ~(FILE_ATTRIBUTE_HIDDEN | FILE_ATTRIBUTE_SYSTEM);
1213         } else if (S_ISLNK(mode)) {
1214                 /* It is good idea that link should be the same type (file/dir) as target */
1215                 fa = FILE_ATTRIBUTE_REPARSE_POINT;
1216
1217                 /*
1218                  * Linux: there are dir/file/symlink and so on.
1219                  * NTFS: symlinks are "dir + reparse" or "file + reparse"
1220                  * It is good idea to create:
1221                  * dir + reparse if 'symname' points to directory
1222                  * or
1223                  * file + reparse if 'symname' points to file
1224                  * Unfortunately kern_path hangs if symname contains 'dir'.
1225                  */
1226
1227                 /*
1228                  *      struct path path;
1229                  *
1230                  *      if (!kern_path(symname, LOOKUP_FOLLOW, &path)){
1231                  *              struct inode *target = d_inode(path.dentry);
1232                  *
1233                  *              if (S_ISDIR(target->i_mode))
1234                  *                      fa |= FILE_ATTRIBUTE_DIRECTORY;
1235                  *              // if ( target->i_sb == sb ){
1236                  *              //      use relative path?
1237                  *              // }
1238                  *              path_put(&path);
1239                  *      }
1240                  */
1241         } else if (S_ISREG(mode)) {
1242                 if (sbi->options->sparse) {
1243                         /* Sparsed regular file, cause option 'sparse'. */
1244                         fa = FILE_ATTRIBUTE_SPARSE_FILE |
1245                              FILE_ATTRIBUTE_ARCHIVE;
1246                 } else if (dir_ni->std_fa & FILE_ATTRIBUTE_COMPRESSED) {
1247                         /* Compressed regular file, if parent is compressed. */
1248                         fa = FILE_ATTRIBUTE_COMPRESSED | FILE_ATTRIBUTE_ARCHIVE;
1249                 } else {
1250                         /* Regular file, default attributes. */
1251                         fa = FILE_ATTRIBUTE_ARCHIVE;
1252                 }
1253         } else {
1254                 fa = FILE_ATTRIBUTE_ARCHIVE;
1255         }
1256
1257         if (!(mode & 0222))
1258                 fa |= FILE_ATTRIBUTE_READONLY;
1259
1260         /* Allocate PATH_MAX bytes. */
1261         new_de = __getname();
1262         if (!new_de) {
1263                 err = -ENOMEM;
1264                 goto out1;
1265         }
1266
1267         /* Mark rw ntfs as dirty. it will be cleared at umount. */
1268         ntfs_set_state(sbi, NTFS_DIRTY_DIRTY);
1269
1270         /* Step 1: allocate and fill new mft record. */
1271         err = ntfs_look_free_mft(sbi, &ino, false, NULL, NULL);
1272         if (err)
1273                 goto out2;
1274
1275         ni = ntfs_new_inode(sbi, ino, fa & FILE_ATTRIBUTE_DIRECTORY);
1276         if (IS_ERR(ni)) {
1277                 err = PTR_ERR(ni);
1278                 ni = NULL;
1279                 goto out3;
1280         }
1281         inode = &ni->vfs_inode;
1282         inode_init_owner(mnt_userns, inode, dir, mode);
1283         mode = inode->i_mode;
1284
1285         inode->i_atime = inode->i_mtime = inode->i_ctime = ni->i_crtime =
1286                 current_time(inode);
1287
1288         rec = ni->mi.mrec;
1289         rec->hard_links = cpu_to_le16(1);
1290         attr = Add2Ptr(rec, le16_to_cpu(rec->attr_off));
1291
1292         /* Get default security id. */
1293         sd = s_default_security;
1294         sd_size = sizeof(s_default_security);
1295
1296         if (is_ntfs3(sbi)) {
1297                 security_id = dir_ni->std_security_id;
1298                 if (le32_to_cpu(security_id) < SECURITY_ID_FIRST) {
1299                         security_id = sbi->security.def_security_id;
1300
1301                         if (security_id == SECURITY_ID_INVALID &&
1302                             !ntfs_insert_security(sbi, sd, sd_size,
1303                                                   &security_id, NULL))
1304                                 sbi->security.def_security_id = security_id;
1305                 }
1306         }
1307
1308         /* Insert standard info. */
1309         std5 = Add2Ptr(attr, SIZEOF_RESIDENT);
1310
1311         if (security_id == SECURITY_ID_INVALID) {
1312                 dsize = sizeof(struct ATTR_STD_INFO);
1313         } else {
1314                 dsize = sizeof(struct ATTR_STD_INFO5);
1315                 std5->security_id = security_id;
1316                 ni->std_security_id = security_id;
1317         }
1318         asize = SIZEOF_RESIDENT + dsize;
1319
1320         attr->type = ATTR_STD;
1321         attr->size = cpu_to_le32(asize);
1322         attr->id = cpu_to_le16(aid++);
1323         attr->res.data_off = SIZEOF_RESIDENT_LE;
1324         attr->res.data_size = cpu_to_le32(dsize);
1325
1326         std5->cr_time = std5->m_time = std5->c_time = std5->a_time =
1327                 kernel2nt(&inode->i_atime);
1328
1329         ni->std_fa = fa;
1330         std5->fa = fa;
1331
1332         attr = Add2Ptr(attr, asize);
1333
1334         /* Insert file name. */
1335         err = fill_name_de(sbi, new_de, name, uni);
1336         if (err)
1337                 goto out4;
1338
1339         mi_get_ref(&ni->mi, &new_de->ref);
1340
1341         fname = (struct ATTR_FILE_NAME *)(new_de + 1);
1342         mi_get_ref(&dir_ni->mi, &fname->home);
1343         fname->dup.cr_time = fname->dup.m_time = fname->dup.c_time =
1344                 fname->dup.a_time = std5->cr_time;
1345         fname->dup.alloc_size = fname->dup.data_size = 0;
1346         fname->dup.fa = std5->fa;
1347         fname->dup.ea_size = fname->dup.reparse = 0;
1348
1349         dsize = le16_to_cpu(new_de->key_size);
1350         asize = ALIGN(SIZEOF_RESIDENT + dsize, 8);
1351
1352         attr->type = ATTR_NAME;
1353         attr->size = cpu_to_le32(asize);
1354         attr->res.data_off = SIZEOF_RESIDENT_LE;
1355         attr->res.flags = RESIDENT_FLAG_INDEXED;
1356         attr->id = cpu_to_le16(aid++);
1357         attr->res.data_size = cpu_to_le32(dsize);
1358         memcpy(Add2Ptr(attr, SIZEOF_RESIDENT), fname, dsize);
1359
1360         attr = Add2Ptr(attr, asize);
1361
1362         if (security_id == SECURITY_ID_INVALID) {
1363                 /* Insert security attribute. */
1364                 asize = SIZEOF_RESIDENT + ALIGN(sd_size, 8);
1365
1366                 attr->type = ATTR_SECURE;
1367                 attr->size = cpu_to_le32(asize);
1368                 attr->id = cpu_to_le16(aid++);
1369                 attr->res.data_off = SIZEOF_RESIDENT_LE;
1370                 attr->res.data_size = cpu_to_le32(sd_size);
1371                 memcpy(Add2Ptr(attr, SIZEOF_RESIDENT), sd, sd_size);
1372
1373                 attr = Add2Ptr(attr, asize);
1374         }
1375
1376         attr->id = cpu_to_le16(aid++);
1377         if (fa & FILE_ATTRIBUTE_DIRECTORY) {
1378                 /*
1379                  * Regular directory or symlink to directory.
1380                  * Create root attribute.
1381                  */
1382                 dsize = sizeof(struct INDEX_ROOT) + sizeof(struct NTFS_DE);
1383                 asize = sizeof(I30_NAME) + SIZEOF_RESIDENT + dsize;
1384
1385                 attr->type = ATTR_ROOT;
1386                 attr->size = cpu_to_le32(asize);
1387
1388                 attr->name_len = ARRAY_SIZE(I30_NAME);
1389                 attr->name_off = SIZEOF_RESIDENT_LE;
1390                 attr->res.data_off =
1391                         cpu_to_le16(sizeof(I30_NAME) + SIZEOF_RESIDENT);
1392                 attr->res.data_size = cpu_to_le32(dsize);
1393                 memcpy(Add2Ptr(attr, SIZEOF_RESIDENT), I30_NAME,
1394                        sizeof(I30_NAME));
1395
1396                 root = Add2Ptr(attr, sizeof(I30_NAME) + SIZEOF_RESIDENT);
1397                 memcpy(root, dir_root, offsetof(struct INDEX_ROOT, ihdr));
1398                 root->ihdr.de_off =
1399                         cpu_to_le32(sizeof(struct INDEX_HDR)); // 0x10
1400                 root->ihdr.used = cpu_to_le32(sizeof(struct INDEX_HDR) +
1401                                               sizeof(struct NTFS_DE));
1402                 root->ihdr.total = root->ihdr.used;
1403
1404                 e = Add2Ptr(root, sizeof(struct INDEX_ROOT));
1405                 e->size = cpu_to_le16(sizeof(struct NTFS_DE));
1406                 e->flags = NTFS_IE_LAST;
1407         } else if (S_ISLNK(mode)) {
1408                 /*
1409                  * Symlink to file.
1410                  * Create empty resident data attribute.
1411                  */
1412                 asize = SIZEOF_RESIDENT;
1413
1414                 /* Insert empty ATTR_DATA */
1415                 attr->type = ATTR_DATA;
1416                 attr->size = cpu_to_le32(SIZEOF_RESIDENT);
1417                 attr->name_off = SIZEOF_RESIDENT_LE;
1418                 attr->res.data_off = SIZEOF_RESIDENT_LE;
1419         } else if (S_ISREG(mode)) {
1420                 /*
1421                  * Regular file. Create empty non resident data attribute.
1422                  */
1423                 attr->type = ATTR_DATA;
1424                 attr->non_res = 1;
1425                 attr->nres.evcn = cpu_to_le64(-1ll);
1426                 if (fa & FILE_ATTRIBUTE_SPARSE_FILE) {
1427                         attr->size = cpu_to_le32(SIZEOF_NONRESIDENT_EX + 8);
1428                         attr->name_off = SIZEOF_NONRESIDENT_EX_LE;
1429                         attr->flags = ATTR_FLAG_SPARSED;
1430                         asize = SIZEOF_NONRESIDENT_EX + 8;
1431                 } else if (fa & FILE_ATTRIBUTE_COMPRESSED) {
1432                         attr->size = cpu_to_le32(SIZEOF_NONRESIDENT_EX + 8);
1433                         attr->name_off = SIZEOF_NONRESIDENT_EX_LE;
1434                         attr->flags = ATTR_FLAG_COMPRESSED;
1435                         attr->nres.c_unit = COMPRESSION_UNIT;
1436                         asize = SIZEOF_NONRESIDENT_EX + 8;
1437                 } else {
1438                         attr->size = cpu_to_le32(SIZEOF_NONRESIDENT + 8);
1439                         attr->name_off = SIZEOF_NONRESIDENT_LE;
1440                         asize = SIZEOF_NONRESIDENT + 8;
1441                 }
1442                 attr->nres.run_off = attr->name_off;
1443         } else {
1444                 /*
1445                  * Node. Create empty resident data attribute.
1446                  */
1447                 attr->type = ATTR_DATA;
1448                 attr->size = cpu_to_le32(SIZEOF_RESIDENT);
1449                 attr->name_off = SIZEOF_RESIDENT_LE;
1450                 if (fa & FILE_ATTRIBUTE_SPARSE_FILE)
1451                         attr->flags = ATTR_FLAG_SPARSED;
1452                 else if (fa & FILE_ATTRIBUTE_COMPRESSED)
1453                         attr->flags = ATTR_FLAG_COMPRESSED;
1454                 attr->res.data_off = SIZEOF_RESIDENT_LE;
1455                 asize = SIZEOF_RESIDENT;
1456                 ni->ni_flags |= NI_FLAG_RESIDENT;
1457         }
1458
1459         if (S_ISDIR(mode)) {
1460                 ni->ni_flags |= NI_FLAG_DIR;
1461                 err = indx_init(&ni->dir, sbi, attr, INDEX_MUTEX_I30);
1462                 if (err)
1463                         goto out4;
1464         } else if (S_ISLNK(mode)) {
1465                 rp = ntfs_create_reparse_buffer(sbi, symname, size, &nsize);
1466
1467                 if (IS_ERR(rp)) {
1468                         err = PTR_ERR(rp);
1469                         rp = NULL;
1470                         goto out4;
1471                 }
1472
1473                 /*
1474                  * Insert ATTR_REPARSE.
1475                  */
1476                 attr = Add2Ptr(attr, asize);
1477                 attr->type = ATTR_REPARSE;
1478                 attr->id = cpu_to_le16(aid++);
1479
1480                 /* Resident or non resident? */
1481                 asize = ALIGN(SIZEOF_RESIDENT + nsize, 8);
1482                 t16 = PtrOffset(rec, attr);
1483
1484                 /*
1485                  * Below function 'ntfs_save_wsl_perm' requires 0x78 bytes.
1486                  * It is good idea to keep extened attributes resident.
1487                  */
1488                 if (asize + t16 + 0x78 + 8 > sbi->record_size) {
1489                         CLST alen;
1490                         CLST clst = bytes_to_cluster(sbi, nsize);
1491
1492                         /* Bytes per runs. */
1493                         t16 = sbi->record_size - t16 - SIZEOF_NONRESIDENT;
1494
1495                         attr->non_res = 1;
1496                         attr->nres.evcn = cpu_to_le64(clst - 1);
1497                         attr->name_off = SIZEOF_NONRESIDENT_LE;
1498                         attr->nres.run_off = attr->name_off;
1499                         attr->nres.data_size = cpu_to_le64(nsize);
1500                         attr->nres.valid_size = attr->nres.data_size;
1501                         attr->nres.alloc_size =
1502                                 cpu_to_le64(ntfs_up_cluster(sbi, nsize));
1503
1504                         err = attr_allocate_clusters(sbi, &ni->file.run, 0, 0,
1505                                                      clst, NULL, 0, &alen, 0,
1506                                                      NULL);
1507                         if (err)
1508                                 goto out5;
1509
1510                         err = run_pack(&ni->file.run, 0, clst,
1511                                        Add2Ptr(attr, SIZEOF_NONRESIDENT), t16,
1512                                        &vcn);
1513                         if (err < 0)
1514                                 goto out5;
1515
1516                         if (vcn != clst) {
1517                                 err = -EINVAL;
1518                                 goto out5;
1519                         }
1520
1521                         asize = SIZEOF_NONRESIDENT + ALIGN(err, 8);
1522                 } else {
1523                         attr->res.data_off = SIZEOF_RESIDENT_LE;
1524                         attr->res.data_size = cpu_to_le32(nsize);
1525                         memcpy(Add2Ptr(attr, SIZEOF_RESIDENT), rp, nsize);
1526                         nsize = 0;
1527                 }
1528                 /* Size of symlink equals the length of input string. */
1529                 inode->i_size = size;
1530
1531                 attr->size = cpu_to_le32(asize);
1532
1533                 err = ntfs_insert_reparse(sbi, IO_REPARSE_TAG_SYMLINK,
1534                                           &new_de->ref);
1535                 if (err)
1536                         goto out5;
1537
1538                 rp_inserted = true;
1539         }
1540
1541         attr = Add2Ptr(attr, asize);
1542         attr->type = ATTR_END;
1543
1544         rec->used = cpu_to_le32(PtrOffset(rec, attr) + 8);
1545         rec->next_attr_id = cpu_to_le16(aid);
1546
1547         /* Step 2: Add new name in index. */
1548         err = indx_insert_entry(&dir_ni->dir, dir_ni, new_de, sbi, fnd, 0);
1549         if (err)
1550                 goto out6;
1551
1552         /* Unlock parent directory before ntfs_init_acl. */
1553         ni_unlock(dir_ni);
1554
1555         inode->i_generation = le16_to_cpu(rec->seq);
1556
1557         dir->i_mtime = dir->i_ctime = inode->i_atime;
1558
1559         if (S_ISDIR(mode)) {
1560                 inode->i_op = &ntfs_dir_inode_operations;
1561                 inode->i_fop = &ntfs_dir_operations;
1562         } else if (S_ISLNK(mode)) {
1563                 inode->i_op = &ntfs_link_inode_operations;
1564                 inode->i_fop = NULL;
1565                 inode->i_mapping->a_ops = &ntfs_aops;
1566                 inode->i_size = size;
1567                 inode_nohighmem(inode);
1568         } else if (S_ISREG(mode)) {
1569                 inode->i_op = &ntfs_file_inode_operations;
1570                 inode->i_fop = &ntfs_file_operations;
1571                 inode->i_mapping->a_ops =
1572                         is_compressed(ni) ? &ntfs_aops_cmpr : &ntfs_aops;
1573                 init_rwsem(&ni->file.run_lock);
1574         } else {
1575                 inode->i_op = &ntfs_special_inode_operations;
1576                 init_special_inode(inode, mode, dev);
1577         }
1578
1579 #ifdef CONFIG_NTFS3_FS_POSIX_ACL
1580         if (!S_ISLNK(mode) && (sb->s_flags & SB_POSIXACL)) {
1581                 err = ntfs_init_acl(mnt_userns, inode, dir);
1582                 if (err)
1583                         goto out7;
1584         } else
1585 #endif
1586         {
1587                 inode->i_flags |= S_NOSEC;
1588         }
1589
1590         /* Write non resident data. */
1591         if (nsize) {
1592                 err = ntfs_sb_write_run(sbi, &ni->file.run, 0, rp, nsize, 0);
1593                 if (err)
1594                         goto out7;
1595         }
1596
1597         /*
1598          * Call 'd_instantiate' after inode->i_op is set
1599          * but before finish_open.
1600          */
1601         d_instantiate(dentry, inode);
1602
1603         ntfs_save_wsl_perm(inode);
1604         mark_inode_dirty(dir);
1605         mark_inode_dirty(inode);
1606
1607         /* Normal exit. */
1608         goto out2;
1609
1610 out7:
1611
1612         /* Undo 'indx_insert_entry'. */
1613         ni_lock_dir(dir_ni);
1614         indx_delete_entry(&dir_ni->dir, dir_ni, new_de + 1,
1615                           le16_to_cpu(new_de->key_size), sbi);
1616         /* ni_unlock(dir_ni); will be called later. */
1617 out6:
1618         if (rp_inserted)
1619                 ntfs_remove_reparse(sbi, IO_REPARSE_TAG_SYMLINK, &new_de->ref);
1620
1621 out5:
1622         if (S_ISDIR(mode) || run_is_empty(&ni->file.run))
1623                 goto out4;
1624
1625         run_deallocate(sbi, &ni->file.run, false);
1626
1627 out4:
1628         clear_rec_inuse(rec);
1629         clear_nlink(inode);
1630         ni->mi.dirty = false;
1631         discard_new_inode(inode);
1632 out3:
1633         ntfs_mark_rec_free(sbi, ino, false);
1634
1635 out2:
1636         __putname(new_de);
1637         kfree(rp);
1638
1639 out1:
1640         if (err) {
1641                 ni_unlock(dir_ni);
1642                 return ERR_PTR(err);
1643         }
1644
1645         unlock_new_inode(inode);
1646
1647         return inode;
1648 }
1649
1650 int ntfs_link_inode(struct inode *inode, struct dentry *dentry)
1651 {
1652         int err;
1653         struct ntfs_inode *ni = ntfs_i(inode);
1654         struct ntfs_sb_info *sbi = inode->i_sb->s_fs_info;
1655         struct NTFS_DE *de;
1656
1657         /* Allocate PATH_MAX bytes. */
1658         de = __getname();
1659         if (!de)
1660                 return -ENOMEM;
1661
1662         /* Mark rw ntfs as dirty. It will be cleared at umount. */
1663         ntfs_set_state(sbi, NTFS_DIRTY_DIRTY);
1664
1665         /* Construct 'de'. */
1666         err = fill_name_de(sbi, de, &dentry->d_name, NULL);
1667         if (err)
1668                 goto out;
1669
1670         err = ni_add_name(ntfs_i(d_inode(dentry->d_parent)), ni, de);
1671 out:
1672         __putname(de);
1673         return err;
1674 }
1675
1676 /*
1677  * ntfs_unlink_inode
1678  *
1679  * inode_operations::unlink
1680  * inode_operations::rmdir
1681  */
1682 int ntfs_unlink_inode(struct inode *dir, const struct dentry *dentry)
1683 {
1684         int err;
1685         struct ntfs_sb_info *sbi = dir->i_sb->s_fs_info;
1686         struct inode *inode = d_inode(dentry);
1687         struct ntfs_inode *ni = ntfs_i(inode);
1688         struct ntfs_inode *dir_ni = ntfs_i(dir);
1689         struct NTFS_DE *de, *de2 = NULL;
1690         int undo_remove;
1691
1692         if (ntfs_is_meta_file(sbi, ni->mi.rno))
1693                 return -EINVAL;
1694
1695         /* Allocate PATH_MAX bytes. */
1696         de = __getname();
1697         if (!de)
1698                 return -ENOMEM;
1699
1700         ni_lock(ni);
1701
1702         if (S_ISDIR(inode->i_mode) && !dir_is_empty(inode)) {
1703                 err = -ENOTEMPTY;
1704                 goto out;
1705         }
1706
1707         err = fill_name_de(sbi, de, &dentry->d_name, NULL);
1708         if (err < 0)
1709                 goto out;
1710
1711         undo_remove = 0;
1712         err = ni_remove_name(dir_ni, ni, de, &de2, &undo_remove);
1713
1714         if (!err) {
1715                 drop_nlink(inode);
1716                 dir->i_mtime = dir->i_ctime = current_time(dir);
1717                 mark_inode_dirty(dir);
1718                 inode->i_ctime = dir->i_ctime;
1719                 if (inode->i_nlink)
1720                         mark_inode_dirty(inode);
1721         } else if (!ni_remove_name_undo(dir_ni, ni, de, de2, undo_remove)) {
1722                 _ntfs_bad_inode(inode);
1723         } else {
1724                 if (ni_is_dirty(dir))
1725                         mark_inode_dirty(dir);
1726                 if (ni_is_dirty(inode))
1727                         mark_inode_dirty(inode);
1728         }
1729
1730 out:
1731         ni_unlock(ni);
1732         __putname(de);
1733         return err;
1734 }
1735
1736 void ntfs_evict_inode(struct inode *inode)
1737 {
1738         truncate_inode_pages_final(&inode->i_data);
1739
1740         if (inode->i_nlink)
1741                 _ni_write_inode(inode, inode_needs_sync(inode));
1742
1743         invalidate_inode_buffers(inode);
1744         clear_inode(inode);
1745
1746         ni_clear(ntfs_i(inode));
1747 }
1748
1749 static noinline int ntfs_readlink_hlp(struct inode *inode, char *buffer,
1750                                       int buflen)
1751 {
1752         int i, err = -EINVAL;
1753         struct ntfs_inode *ni = ntfs_i(inode);
1754         struct super_block *sb = inode->i_sb;
1755         struct ntfs_sb_info *sbi = sb->s_fs_info;
1756         u64 size;
1757         u16 ulen = 0;
1758         void *to_free = NULL;
1759         struct REPARSE_DATA_BUFFER *rp;
1760         const __le16 *uname;
1761         struct ATTRIB *attr;
1762
1763         /* Reparse data present. Try to parse it. */
1764         static_assert(!offsetof(struct REPARSE_DATA_BUFFER, ReparseTag));
1765         static_assert(sizeof(u32) == sizeof(rp->ReparseTag));
1766
1767         *buffer = 0;
1768
1769         attr = ni_find_attr(ni, NULL, NULL, ATTR_REPARSE, NULL, 0, NULL, NULL);
1770         if (!attr)
1771                 goto out;
1772
1773         if (!attr->non_res) {
1774                 rp = resident_data_ex(attr, sizeof(struct REPARSE_DATA_BUFFER));
1775                 if (!rp)
1776                         goto out;
1777                 size = le32_to_cpu(attr->res.data_size);
1778         } else {
1779                 size = le64_to_cpu(attr->nres.data_size);
1780                 rp = NULL;
1781         }
1782
1783         if (size > sbi->reparse.max_size || size <= sizeof(u32))
1784                 goto out;
1785
1786         if (!rp) {
1787                 rp = kmalloc(size, GFP_NOFS);
1788                 if (!rp) {
1789                         err = -ENOMEM;
1790                         goto out;
1791                 }
1792                 to_free = rp;
1793                 /* Read into temporal buffer. */
1794                 err = ntfs_read_run_nb(sbi, &ni->file.run, 0, rp, size, NULL);
1795                 if (err)
1796                         goto out;
1797         }
1798
1799         /* Microsoft Tag. */
1800         switch (rp->ReparseTag) {
1801         case IO_REPARSE_TAG_MOUNT_POINT:
1802                 /* Mount points and junctions. */
1803                 /* Can we use 'Rp->MountPointReparseBuffer.PrintNameLength'? */
1804                 if (size <= offsetof(struct REPARSE_DATA_BUFFER,
1805                                      MountPointReparseBuffer.PathBuffer))
1806                         goto out;
1807                 uname = Add2Ptr(rp,
1808                                 offsetof(struct REPARSE_DATA_BUFFER,
1809                                          MountPointReparseBuffer.PathBuffer) +
1810                                         le16_to_cpu(rp->MountPointReparseBuffer
1811                                                             .PrintNameOffset));
1812                 ulen = le16_to_cpu(rp->MountPointReparseBuffer.PrintNameLength);
1813                 break;
1814
1815         case IO_REPARSE_TAG_SYMLINK:
1816                 /* FolderSymbolicLink */
1817                 /* Can we use 'Rp->SymbolicLinkReparseBuffer.PrintNameLength'? */
1818                 if (size <= offsetof(struct REPARSE_DATA_BUFFER,
1819                                      SymbolicLinkReparseBuffer.PathBuffer))
1820                         goto out;
1821                 uname = Add2Ptr(
1822                         rp, offsetof(struct REPARSE_DATA_BUFFER,
1823                                      SymbolicLinkReparseBuffer.PathBuffer) +
1824                                     le16_to_cpu(rp->SymbolicLinkReparseBuffer
1825                                                         .PrintNameOffset));
1826                 ulen = le16_to_cpu(
1827                         rp->SymbolicLinkReparseBuffer.PrintNameLength);
1828                 break;
1829
1830         case IO_REPARSE_TAG_CLOUD:
1831         case IO_REPARSE_TAG_CLOUD_1:
1832         case IO_REPARSE_TAG_CLOUD_2:
1833         case IO_REPARSE_TAG_CLOUD_3:
1834         case IO_REPARSE_TAG_CLOUD_4:
1835         case IO_REPARSE_TAG_CLOUD_5:
1836         case IO_REPARSE_TAG_CLOUD_6:
1837         case IO_REPARSE_TAG_CLOUD_7:
1838         case IO_REPARSE_TAG_CLOUD_8:
1839         case IO_REPARSE_TAG_CLOUD_9:
1840         case IO_REPARSE_TAG_CLOUD_A:
1841         case IO_REPARSE_TAG_CLOUD_B:
1842         case IO_REPARSE_TAG_CLOUD_C:
1843         case IO_REPARSE_TAG_CLOUD_D:
1844         case IO_REPARSE_TAG_CLOUD_E:
1845         case IO_REPARSE_TAG_CLOUD_F:
1846                 err = sizeof("OneDrive") - 1;
1847                 if (err > buflen)
1848                         err = buflen;
1849                 memcpy(buffer, "OneDrive", err);
1850                 goto out;
1851
1852         default:
1853                 if (IsReparseTagMicrosoft(rp->ReparseTag)) {
1854                         /* Unknown Microsoft Tag. */
1855                         goto out;
1856                 }
1857                 if (!IsReparseTagNameSurrogate(rp->ReparseTag) ||
1858                     size <= sizeof(struct REPARSE_POINT)) {
1859                         goto out;
1860                 }
1861
1862                 /* Users tag. */
1863                 uname = Add2Ptr(rp, sizeof(struct REPARSE_POINT));
1864                 ulen = le16_to_cpu(rp->ReparseDataLength) -
1865                        sizeof(struct REPARSE_POINT);
1866         }
1867
1868         /* Convert nlen from bytes to UNICODE chars. */
1869         ulen >>= 1;
1870
1871         /* Check that name is available. */
1872         if (!ulen || uname + ulen > (__le16 *)Add2Ptr(rp, size))
1873                 goto out;
1874
1875         /* If name is already zero terminated then truncate it now. */
1876         if (!uname[ulen - 1])
1877                 ulen -= 1;
1878
1879         err = ntfs_utf16_to_nls(sbi, uname, ulen, buffer, buflen);
1880
1881         if (err < 0)
1882                 goto out;
1883
1884         /* Translate Windows '\' into Linux '/'. */
1885         for (i = 0; i < err; i++) {
1886                 if (buffer[i] == '\\')
1887                         buffer[i] = '/';
1888         }
1889
1890         /* Always set last zero. */
1891         buffer[err] = 0;
1892 out:
1893         kfree(to_free);
1894         return err;
1895 }
1896
1897 static const char *ntfs_get_link(struct dentry *de, struct inode *inode,
1898                                  struct delayed_call *done)
1899 {
1900         int err;
1901         char *ret;
1902
1903         if (!de)
1904                 return ERR_PTR(-ECHILD);
1905
1906         ret = kmalloc(PAGE_SIZE, GFP_NOFS);
1907         if (!ret)
1908                 return ERR_PTR(-ENOMEM);
1909
1910         err = ntfs_readlink_hlp(inode, ret, PAGE_SIZE);
1911         if (err < 0) {
1912                 kfree(ret);
1913                 return ERR_PTR(err);
1914         }
1915
1916         set_delayed_call(done, kfree_link, ret);
1917
1918         return ret;
1919 }
1920
1921 // clang-format off
1922 const struct inode_operations ntfs_link_inode_operations = {
1923         .get_link       = ntfs_get_link,
1924         .setattr        = ntfs3_setattr,
1925         .listxattr      = ntfs_listxattr,
1926         .permission     = ntfs_permission,
1927 };
1928
1929 const struct address_space_operations ntfs_aops = {
1930         .read_folio     = ntfs_read_folio,
1931         .readahead      = ntfs_readahead,
1932         .writepage      = ntfs_writepage,
1933         .writepages     = ntfs_writepages,
1934         .write_begin    = ntfs_write_begin,
1935         .write_end      = ntfs_write_end,
1936         .direct_IO      = ntfs_direct_IO,
1937         .bmap           = ntfs_bmap,
1938         .dirty_folio    = block_dirty_folio,
1939         .invalidate_folio = block_invalidate_folio,
1940 };
1941
1942 const struct address_space_operations ntfs_aops_cmpr = {
1943         .read_folio     = ntfs_read_folio,
1944         .readahead      = ntfs_readahead,
1945 };
1946 // clang-format on